Proposal

In current versions of Prometheus, there is a metric scrape_duration_seconds that shows the end to end scrape time. However, when debugging slow scrapes, it's hard to know if the issue was in the target, the network between Prometheus and the target, or in Prometheus. I would like to propose that even if we do not add more metrics, we add debug level logs to track the durations of the following:

1. How long it takes to send an HTTP request and receive a response to /metrics from the target
2. How long it takes to parse the response
3. How long it takes to ingest the data in the response